关于 decoder_outputs[:,t,:] = decoder_output_t torch.topk, torch.max(),torch.argmax()的演示

Posted 2021-09-03 ZSYL

关于 PyTorch 特殊方法的演示

• 1. decoder_outputs[:,t,:] = decoder_output_t
• 2. torch.topk, torch.max(), torch.argmax()
• 3. unsqueeze()

1. decoder_outputs[:,t,:] = decoder_output_t

decoder_outputs 形状 [batch_size, seq_len, vocab_size]
decoder_output_t 形状[batch_size, vocab_size]

示例代码：

import torch
 
a = torch.zeros((2, 3, 5))
print(a.size())
print(a)
 
b = torch.randn((2, 5))
print(b.size())
print(b)
 
a[:, 0, :] = b
print(a.size())
print(a)

运行结果：

2. torch.topk, torch.max(), torch.argmax()

value, index = torch.topk(decoder_output_t , k = 1)
decoder_output_t [batch_size, vocab_size]

示例代码：

import torch
 
a = torch.randn((3, 5))
print(a.size())
print(a)
 
values, index = torch.topk(a, k=1)
print(values)
print(index)
print(index.size())
 
values, index = torch.max(a, dim=-1)
print(values)
print(index)
print(index.size())
 
index = torch.argmax(a, dim=-1)
print(index)
print(index.size())
 
index = a.argmax(dim=-1)
print(index)
print(index.size())

运行结果：

3. unsqueeze()

若使用 teacher forcing ,将采用下次真实值作为下个time step的输入

# 注意unsqueeze 相当于浅拷贝,不会对原张量进行修改
decoder_input = target[:,t].unsqueeze(-1)
target 形状 [batch_size, seq_len]
decoder_input 要求形状[batch_size, 1]

示例代码：

import torch
 
a = torch.randn((3, 5))
print(a.size())
print(a)
 
b = a[:, 3]
print(b.size())
print(b)
c = b.unsqueeze(-1)
print(c.size())
print(c)

运行结果：

参考：link